RE: [boost] rational.hpp - abs() again

From: Moore, Paul (paul.moore_at_[hidden])
Date: 2001-01-24 09:33:54

From: Gabriel Dos Reis [mailto:Gabriel.Dos-Reis_at_[hidden]]
> That is not true.
[...]
> I'am mainly concerned with your comment about std::abs.  It was not
> accurate.

I apologise. The workaround in rational.hpp was put in a long time ago,
based (if I recall correctly) on a report + fix from someone of breakage.
The comment alongside the fix came from that time. I don't have the
compiler(s) to test things like this (at the time, I didn't even have mingw
working) so I have to rely on others' reports.

This does raise an issue. If workarounds for broken compilers are required,
the maintainer of a piece of code may not have any way of testing whether
the suggested workaround is needed or correct. I have always worked on the
basis of accepting any suggestions for compilers I don't use (essentially
everything but MSVC) without question. When it comes to making changes, I'm
then faced with the need to either remove the workaround (and hope that
someone will find and re-fix the issue) or attempt to modify the workaround
to do the right thing with the code changes I have made. Neither is a
particularly good option.

There's also the issue of when/how workarounds should be removed as
compilers improve. (Not relevant here, as it appears that gcc may never have
been broken...)

Paul.